Next up in my Srull & Wyer tracing, is Herr, Sherman & Fazios paper, ”on the consequences of priming: Assimilation and contrast effects”, published 1983 in the Journal of experimental social psychology.*
The research here is more conceptually related than directly pushing further on either the Donald story, or using sentence unscrambling tasks.
Where they are similar is that the work investigates incidental priming (ferocity or size in this case), and how this prime then potentially influences the ratings of ambiguous stimuli (non-existing animals).
Where they extend is that they are checking whether priming always leads to assimilation (which one can argue is the case in the original Srull & Wyer, as well as in the Bargh & Pietromonaco article), or if there are instances where the prime leads to a contrast effect instead.
Some notes first: I had trouble getting my head around the results, until I graphed it in a way I could understand – and then things became much clearer. I’ll see if I can share those graphs.
The paper sports two allegedly significant 3-way interactions, where one has a p-levels of .058, and the other .067. We don’t know if those p-values are approaching or running away from that arbitrary cut-off level of significance, as the authors don’t say. No matter. It was 1983, and those p-values like to boogie anyway.
The puzzle they try to answer (according to the introduction) is that whereas the priming work shows an assimilation effect (judgment gets drawn closer to the accessible category), it is much more common in the social judgment literature to see a contrast effect. Perhaps, they think, it is the ambiguity of the stimulus-to-be-judged. For non-ambigous stimuli, a contrast effect might emerge. In addition, the extremity of the prime may matter. If the primed category is very extreme, perhaps there will be no assimilation, even of ambiguous stimuli. It is simply too far away to serve as a reasonable category, and thus a contrast effect will be evident. They cite plenty of older evidence for this.
They don’t evaluate this hypothesis in the domain of hostility or kindness, though. Instead they use judgment of features of animals that are either known or unknown.
The prediction would then be that for known animals, priming will simply result in the judgment being contrasted away from the activated category. For the unknown animals, priming with extreme exemplars will also result in contrast, whereas priming with moderate exemplars will result in assimilation. I will show a table of predictions below.
The set up.
In the first experiment, they test rating of ferocity, and in the second rating of size.
First, they gather together 20 animals as exemplars of ferocity. Four are rated as extremely ferocious (Grizzly bear, Tiger, Lion, Shark) Four are rated as moderately ferocious (Vulture, Wolf, Rhinoceros, Badger). Then we have four moderately non-ferocious (Kangaroo, Opossum, Cat, Seal), and finally the meekest of all (Dove, Kitten, Rabbit Puppy).
There is also the 4 remaining – the moderate set: Fox, Porcupine, Weasel and Bat. These are the animals that will later be rated on ferocity, along with two imagninary animals,, the Jabo and the Lemphor.
The priming task is an ostensible Stroop task. Participants see a series of pairs of slides. (10 pairs in total). The first slide in each pair shows a word, that the participant is supposed to memorize. After a while, they see the second slide. For example RED . Now they are supposed to say the memorized word, and the color of the ink as fast as possible. The researcher is standing there with a stop-watch, tick-tock.
Most words are neutral, but in the 3rd, 5th, 7th and 8th position they slip in the animal name.
The task is modeled on the Higgins, Rholes and Jones (1977) task. (This seems to be one of the Ur-papers for this particular line of questioning).
This ends up being 4 priming conditions, 20 people in each cell (better than 8). The conditions can be divided into two dimensions: Ferocious or not, Extreme or Moderate.
This is where I got confused, as extremity bundles together the kitten and the lion, but I think I got the understanding sorted in the end.
Next, in good Social Psychology tradition, the students are lied to. It is the standard, “two experiment” set-up, where , oh, by the way, we have a short other experiment that we would be ever so grateful if you took part in. Won’t take long, promise.
This is the rating task. They have cards with the animals (I’m assuming they give the name. Who has ever seen a Lemphor?), and ratings sheet. For each animal, they give a rating, on an 11 point scale, how 1) ferocious the animal is, 2) the “likelihood that the animal would cause harm” and 3) “the seriousness of harm the animal could inflict”.
Half of the participants rate the real animals before the imaginary, and the other half the imaginary before the real.
I note there are 4 real animals and 2 fake, but the mean ratings for all of them are thrown into the ANOVA later. Hmmmm.
Then they debrief, and nobody believed that the two experiments really were related.
First, something about how they created the scores. First they averaged each rating across the animals (so a mean score of ferocity across 4 real animals, and a mean score of ferocity across 2 imagniary animals). This results in 3 means. These means were then added up to a composite score.
I was a little bit surprised fining scores above 11 on an 11 point likert scale, but this explains it. I think I would have preferred averages of averages (to relate it better to the scale), but these are just linear combinations, so no matter.
As promised, this is the pattern of ratings that we would expect:
|Low ferocious prime||High Ferocios prime|
And, what do they find? Let me give the actual ratings.
|Low Ferocious prime||High Ferocious prime|
The ambiguous animals are, overall, rated as more ferocious than the real animals. This is also significant in the ANOVA: Then, there’s the allgeded 3-way interaction between type of animal (real or imaginary) Ferocity (Meek or ferocious) and extremity (Extreme or moderate).
They seem to take this as a reason to do a planned contrast to test the above prediction (that only the ambiguous animals get assimilated in the moderate prim condition, whereas the rest get contrasted away). This one is significant, F(1,72) = 6,50 p < .01. (I don’t know how they set this together, though. I’ve spent quite a bit puzzling this out.).
If you look at it, the one that sticks out (and they mention it too) is the rating of the ambiguous animals after having been exposed to a moderately ferocious prime. This is by far the highest composite ferocity score.
You can look at the above again for either the real or the ambiguous animals (and this is what they do analyze). You would expect that the real animals would be rated as more ferocious after being exposed to the low ferocious prime, and less ferocious after being exposed to the high ferocious prime (contrast effect). Just sheer eye-balling says that this is not there, and neither does their ANOVA.
For the ambiguous, one would expect higher ferocity after extreme low ferocious prime than after extreme high ferocious prim (contrast) which eyeballing may suggest is weakly the case. Also, one would think that (according to assimilation) the rating would be of lower ferocity after moderate low ferocious prime, and higher ferocity after higher ferocious prime. Which you kinda see. Though the interaction is not significant.
Like they say, the patterns are in the expected direction, but not statistically significant. Well, good enough for publishing in 1983.
On to experiment 2.
Set-up is the same, but there are a few differences:
First, they double the amount of participants, from 80 to 160 (40 per cell). Yay. More power!
Second, instead of priming ferocity, they prime size. Like in the first experiment, they test up a number of animals, as follows:
- Large: Whale, Elephant, Hippo, Rhinoceros.
- Moderately large: Antelope, Cow, Lion, Tiger
- Moderate: Wolf, Sheep, Pig, Goat
- Moderately small: Porcupine, Gopher, Groundhog, Cat.
- Small: Snail, Flea, Minnow, Ant.
As in the first experiment, the Middle category were rated, whereas the other categories were used in the priming task.
Here are the results.
|Low size prime||High Size prime|
Eyeballing the ratings of the real animals, they don’t seem to really differ much. They are actually always rated as larger than the ambiguous animals. But, their analysis also shows that the low-size prime results in larger ratings of the real animals, than the high size prime.
For the unreal animals there is the predicted interaction. The moderate primes seem to result in assimilation, whereas the extreme prime results in contrast, and this interaction is significant: F(1,152) = 6,70, p = .011
Yay for more power!
Herr, Paul M, Sherman, Steven J., & Fazio, Russell H. (1983) On the consequences of priming: Assimilation and contrast effects. Journal of Experimental Social Psychology, 19, 323-340
*Jim and Russ were professors in my department when I was going to graduate school, although this was published long before I ever thought of doing a PhD in anything anywhere.